1 use crate::lexer::unicode_chars::UNICODE_ARRAY;
2 use rustc_ast::ast::{self, AttrStyle};
3 use rustc_ast::token::{self, CommentKind, Token, TokenKind};
4 use rustc_ast::tokenstream::{Spacing, TokenStream};
5 use rustc_ast::util::unicode::contains_text_flow_control_chars;
6 use rustc_errors::{error_code, Applicability, DiagnosticBuilder, FatalError, PResult};
7 use rustc_lexer::unescape::{self, Mode};
8 use rustc_lexer::{Base, DocStyle, RawStrError};
9 use rustc_session::lint::builtin::{
10 RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX, TEXT_DIRECTION_CODEPOINT_IN_COMMENT,
11 };
12 use rustc_session::lint::BuiltinLintDiagnostics;
13 use rustc_session::parse::ParseSess;
14 use rustc_span::symbol::{sym, Symbol};
15 use rustc_span::{edition::Edition, BytePos, Pos, Span};
16
17 use tracing::debug;
18
19 mod tokentrees;
20 mod unescape_error_reporting;
21 mod unicode_chars;
22
23 use unescape_error_reporting::{emit_unescape_error, escaped_char};
24
25 #[derive(Clone, Debug)]
26 pub struct UnmatchedBrace {
27 pub expected_delim: token::DelimToken,
28 pub found_delim: Option<token::DelimToken>,
29 pub found_span: Span,
30 pub unclosed_span: Option<Span>,
31 pub candidate_span: Option<Span>,
32 }
33
parse_token_trees<'a>( sess: &'a ParseSess, src: &'a str, start_pos: BytePos, override_span: Option<Span>, ) -> (PResult<'a, TokenStream>, Vec<UnmatchedBrace>)34 crate fn parse_token_trees<'a>(
35 sess: &'a ParseSess,
36 src: &'a str,
37 start_pos: BytePos,
38 override_span: Option<Span>,
39 ) -> (PResult<'a, TokenStream>, Vec<UnmatchedBrace>) {
40 StringReader { sess, start_pos, pos: start_pos, end_src_index: src.len(), src, override_span }
41 .into_token_trees()
42 }
43
44 struct StringReader<'a> {
45 sess: &'a ParseSess,
46 /// Initial position, read-only.
47 start_pos: BytePos,
48 /// The absolute offset within the source_map of the current character.
49 pos: BytePos,
50 /// Stop reading src at this index.
51 end_src_index: usize,
52 /// Source text to tokenize.
53 src: &'a str,
54 override_span: Option<Span>,
55 }
56
57 impl<'a> StringReader<'a> {
mk_sp(&self, lo: BytePos, hi: BytePos) -> Span58 fn mk_sp(&self, lo: BytePos, hi: BytePos) -> Span {
59 self.override_span.unwrap_or_else(|| Span::with_root_ctxt(lo, hi))
60 }
61
62 /// Returns the next token, and info about preceding whitespace, if any.
next_token(&mut self) -> (Spacing, Token)63 fn next_token(&mut self) -> (Spacing, Token) {
64 let mut spacing = Spacing::Joint;
65
66 // Skip `#!` at the start of the file
67 let start_src_index = self.src_index(self.pos);
68 let text: &str = &self.src[start_src_index..self.end_src_index];
69 let is_beginning_of_file = self.pos == self.start_pos;
70 if is_beginning_of_file {
71 if let Some(shebang_len) = rustc_lexer::strip_shebang(text) {
72 self.pos = self.pos + BytePos::from_usize(shebang_len);
73 spacing = Spacing::Alone;
74 }
75 }
76
77 // Skip trivial (whitespace & comments) tokens
78 loop {
79 let start_src_index = self.src_index(self.pos);
80 let text: &str = &self.src[start_src_index..self.end_src_index];
81
82 if text.is_empty() {
83 let span = self.mk_sp(self.pos, self.pos);
84 return (spacing, Token::new(token::Eof, span));
85 }
86
87 let token = rustc_lexer::first_token(text);
88
89 let start = self.pos;
90 self.pos = self.pos + BytePos::from_usize(token.len);
91
92 debug!("next_token: {:?}({:?})", token.kind, self.str_from(start));
93
94 match self.cook_lexer_token(token.kind, start) {
95 Some(kind) => {
96 let span = self.mk_sp(start, self.pos);
97 return (spacing, Token::new(kind, span));
98 }
99 None => spacing = Spacing::Alone,
100 }
101 }
102 }
103
104 /// Report a fatal lexical error with a given span.
fatal_span(&self, sp: Span, m: &str) -> FatalError105 fn fatal_span(&self, sp: Span, m: &str) -> FatalError {
106 self.sess.span_diagnostic.span_fatal(sp, m)
107 }
108
109 /// Report a lexical error with a given span.
err_span(&self, sp: Span, m: &str)110 fn err_span(&self, sp: Span, m: &str) {
111 self.sess.span_diagnostic.struct_span_err(sp, m).emit();
112 }
113
114 /// Report a fatal error spanning [`from_pos`, `to_pos`).
fatal_span_(&self, from_pos: BytePos, to_pos: BytePos, m: &str) -> FatalError115 fn fatal_span_(&self, from_pos: BytePos, to_pos: BytePos, m: &str) -> FatalError {
116 self.fatal_span(self.mk_sp(from_pos, to_pos), m)
117 }
118
119 /// Report a lexical error spanning [`from_pos`, `to_pos`).
err_span_(&self, from_pos: BytePos, to_pos: BytePos, m: &str)120 fn err_span_(&self, from_pos: BytePos, to_pos: BytePos, m: &str) {
121 self.err_span(self.mk_sp(from_pos, to_pos), m)
122 }
123
struct_fatal_span_char( &self, from_pos: BytePos, to_pos: BytePos, m: &str, c: char, ) -> DiagnosticBuilder<'a>124 fn struct_fatal_span_char(
125 &self,
126 from_pos: BytePos,
127 to_pos: BytePos,
128 m: &str,
129 c: char,
130 ) -> DiagnosticBuilder<'a> {
131 self.sess
132 .span_diagnostic
133 .struct_span_fatal(self.mk_sp(from_pos, to_pos), &format!("{}: {}", m, escaped_char(c)))
134 }
135
136 /// Detect usages of Unicode codepoints changing the direction of the text on screen and loudly
137 /// complain about it.
lint_unicode_text_flow(&self, start: BytePos)138 fn lint_unicode_text_flow(&self, start: BytePos) {
139 // Opening delimiter of the length 2 is not included into the comment text.
140 let content_start = start + BytePos(2);
141 let content = self.str_from(content_start);
142 if contains_text_flow_control_chars(content) {
143 let span = self.mk_sp(start, self.pos);
144 self.sess.buffer_lint_with_diagnostic(
145 &TEXT_DIRECTION_CODEPOINT_IN_COMMENT,
146 span,
147 ast::CRATE_NODE_ID,
148 "unicode codepoint changing visible direction of text present in comment",
149 BuiltinLintDiagnostics::UnicodeTextFlow(span, content.to_string()),
150 );
151 }
152 }
153
154 /// Turns simple `rustc_lexer::TokenKind` enum into a rich
155 /// `rustc_ast::TokenKind`. This turns strings into interned
156 /// symbols and runs additional validation.
cook_lexer_token(&self, token: rustc_lexer::TokenKind, start: BytePos) -> Option<TokenKind>157 fn cook_lexer_token(&self, token: rustc_lexer::TokenKind, start: BytePos) -> Option<TokenKind> {
158 Some(match token {
159 rustc_lexer::TokenKind::LineComment { doc_style } => {
160 // Skip non-doc comments
161 let doc_style = if let Some(doc_style) = doc_style {
162 doc_style
163 } else {
164 self.lint_unicode_text_flow(start);
165 return None;
166 };
167
168 // Opening delimiter of the length 3 is not included into the symbol.
169 let content_start = start + BytePos(3);
170 let content = self.str_from(content_start);
171 self.cook_doc_comment(content_start, content, CommentKind::Line, doc_style)
172 }
173 rustc_lexer::TokenKind::BlockComment { doc_style, terminated } => {
174 if !terminated {
175 let msg = match doc_style {
176 Some(_) => "unterminated block doc-comment",
177 None => "unterminated block comment",
178 };
179 let last_bpos = self.pos;
180 self.sess.span_diagnostic.span_fatal_with_code(
181 self.mk_sp(start, last_bpos),
182 msg,
183 error_code!(E0758),
184 );
185 }
186
187 // Skip non-doc comments
188 let doc_style = if let Some(doc_style) = doc_style {
189 doc_style
190 } else {
191 self.lint_unicode_text_flow(start);
192 return None;
193 };
194
195 // Opening delimiter of the length 3 and closing delimiter of the length 2
196 // are not included into the symbol.
197 let content_start = start + BytePos(3);
198 let content_end = self.pos - BytePos(if terminated { 2 } else { 0 });
199 let content = self.str_from_to(content_start, content_end);
200 self.cook_doc_comment(content_start, content, CommentKind::Block, doc_style)
201 }
202 rustc_lexer::TokenKind::Whitespace => return None,
203 rustc_lexer::TokenKind::Ident
204 | rustc_lexer::TokenKind::RawIdent
205 | rustc_lexer::TokenKind::UnknownPrefix => {
206 let is_raw_ident = token == rustc_lexer::TokenKind::RawIdent;
207 let is_unknown_prefix = token == rustc_lexer::TokenKind::UnknownPrefix;
208 let mut ident_start = start;
209 if is_raw_ident {
210 ident_start = ident_start + BytePos(2);
211 }
212 if is_unknown_prefix {
213 self.report_unknown_prefix(start);
214 }
215 let sym = nfc_normalize(self.str_from(ident_start));
216 let span = self.mk_sp(start, self.pos);
217 self.sess.symbol_gallery.insert(sym, span);
218 if is_raw_ident {
219 if !sym.can_be_raw() {
220 self.err_span(span, &format!("`{}` cannot be a raw identifier", sym));
221 }
222 self.sess.raw_identifier_spans.borrow_mut().push(span);
223 }
224 token::Ident(sym, is_raw_ident)
225 }
226 rustc_lexer::TokenKind::InvalidIdent
227 // Do not recover an identifier with emoji if the codepoint is a confusable
228 // with a recoverable substitution token, like `➖`.
229 if UNICODE_ARRAY
230 .iter()
231 .find(|&&(c, _, _)| {
232 let sym = self.str_from(start);
233 sym.chars().count() == 1 && c == sym.chars().next().unwrap()
234 })
235 .is_none() =>
236 {
237 let sym = nfc_normalize(self.str_from(start));
238 let span = self.mk_sp(start, self.pos);
239 self.sess.bad_unicode_identifiers.borrow_mut().entry(sym).or_default().push(span);
240 token::Ident(sym, false)
241 }
242 rustc_lexer::TokenKind::Literal { kind, suffix_start } => {
243 let suffix_start = start + BytePos(suffix_start as u32);
244 let (kind, symbol) = self.cook_lexer_literal(start, suffix_start, kind);
245 let suffix = if suffix_start < self.pos {
246 let string = self.str_from(suffix_start);
247 if string == "_" {
248 self.sess
249 .span_diagnostic
250 .struct_span_warn(
251 self.mk_sp(suffix_start, self.pos),
252 "underscore literal suffix is not allowed",
253 )
254 .warn(
255 "this was previously accepted by the compiler but is \
256 being phased out; it will become a hard error in \
257 a future release!",
258 )
259 .note(
260 "see issue #42326 \
261 <https://github.com/rust-lang/rust/issues/42326> \
262 for more information",
263 )
264 .emit();
265 None
266 } else {
267 Some(Symbol::intern(string))
268 }
269 } else {
270 None
271 };
272 token::Literal(token::Lit { kind, symbol, suffix })
273 }
274 rustc_lexer::TokenKind::Lifetime { starts_with_number } => {
275 // Include the leading `'` in the real identifier, for macro
276 // expansion purposes. See #12512 for the gory details of why
277 // this is necessary.
278 let lifetime_name = self.str_from(start);
279 if starts_with_number {
280 self.err_span_(start, self.pos, "lifetimes cannot start with a number");
281 }
282 let ident = Symbol::intern(lifetime_name);
283 token::Lifetime(ident)
284 }
285 rustc_lexer::TokenKind::Semi => token::Semi,
286 rustc_lexer::TokenKind::Comma => token::Comma,
287 rustc_lexer::TokenKind::Dot => token::Dot,
288 rustc_lexer::TokenKind::OpenParen => token::OpenDelim(token::Paren),
289 rustc_lexer::TokenKind::CloseParen => token::CloseDelim(token::Paren),
290 rustc_lexer::TokenKind::OpenBrace => token::OpenDelim(token::Brace),
291 rustc_lexer::TokenKind::CloseBrace => token::CloseDelim(token::Brace),
292 rustc_lexer::TokenKind::OpenBracket => token::OpenDelim(token::Bracket),
293 rustc_lexer::TokenKind::CloseBracket => token::CloseDelim(token::Bracket),
294 rustc_lexer::TokenKind::At => token::At,
295 rustc_lexer::TokenKind::Pound => token::Pound,
296 rustc_lexer::TokenKind::Tilde => token::Tilde,
297 rustc_lexer::TokenKind::Question => token::Question,
298 rustc_lexer::TokenKind::Colon => token::Colon,
299 rustc_lexer::TokenKind::Dollar => token::Dollar,
300 rustc_lexer::TokenKind::Eq => token::Eq,
301 rustc_lexer::TokenKind::Bang => token::Not,
302 rustc_lexer::TokenKind::Lt => token::Lt,
303 rustc_lexer::TokenKind::Gt => token::Gt,
304 rustc_lexer::TokenKind::Minus => token::BinOp(token::Minus),
305 rustc_lexer::TokenKind::And => token::BinOp(token::And),
306 rustc_lexer::TokenKind::Or => token::BinOp(token::Or),
307 rustc_lexer::TokenKind::Plus => token::BinOp(token::Plus),
308 rustc_lexer::TokenKind::Star => token::BinOp(token::Star),
309 rustc_lexer::TokenKind::Slash => token::BinOp(token::Slash),
310 rustc_lexer::TokenKind::Caret => token::BinOp(token::Caret),
311 rustc_lexer::TokenKind::Percent => token::BinOp(token::Percent),
312
313 rustc_lexer::TokenKind::Unknown | rustc_lexer::TokenKind::InvalidIdent => {
314 let c = self.str_from(start).chars().next().unwrap();
315 let mut err =
316 self.struct_fatal_span_char(start, self.pos, "unknown start of token", c);
317 // FIXME: the lexer could be used to turn the ASCII version of unicode homoglyphs,
318 // instead of keeping a table in `check_for_substitution`into the token. Ideally,
319 // this should be inside `rustc_lexer`. However, we should first remove compound
320 // tokens like `<<` from `rustc_lexer`, and then add fancier error recovery to it,
321 // as there will be less overall work to do this way.
322 let token = unicode_chars::check_for_substitution(self, start, c, &mut err);
323 if c == '\x00' {
324 err.help("source files must contain UTF-8 encoded text, unexpected null bytes might occur when a different encoding is used");
325 }
326 err.emit();
327 token?
328 }
329 })
330 }
331
cook_doc_comment( &self, content_start: BytePos, content: &str, comment_kind: CommentKind, doc_style: DocStyle, ) -> TokenKind332 fn cook_doc_comment(
333 &self,
334 content_start: BytePos,
335 content: &str,
336 comment_kind: CommentKind,
337 doc_style: DocStyle,
338 ) -> TokenKind {
339 if content.contains('\r') {
340 for (idx, _) in content.char_indices().filter(|&(_, c)| c == '\r') {
341 self.err_span_(
342 content_start + BytePos(idx as u32),
343 content_start + BytePos(idx as u32 + 1),
344 match comment_kind {
345 CommentKind::Line => "bare CR not allowed in doc-comment",
346 CommentKind::Block => "bare CR not allowed in block doc-comment",
347 },
348 );
349 }
350 }
351
352 let attr_style = match doc_style {
353 DocStyle::Outer => AttrStyle::Outer,
354 DocStyle::Inner => AttrStyle::Inner,
355 };
356
357 token::DocComment(comment_kind, attr_style, Symbol::intern(content))
358 }
359
cook_lexer_literal( &self, start: BytePos, suffix_start: BytePos, kind: rustc_lexer::LiteralKind, ) -> (token::LitKind, Symbol)360 fn cook_lexer_literal(
361 &self,
362 start: BytePos,
363 suffix_start: BytePos,
364 kind: rustc_lexer::LiteralKind,
365 ) -> (token::LitKind, Symbol) {
366 // prefix means `"` or `br"` or `r###"`, ...
367 let (lit_kind, mode, prefix_len, postfix_len) = match kind {
368 rustc_lexer::LiteralKind::Char { terminated } => {
369 if !terminated {
370 self.sess.span_diagnostic.span_fatal_with_code(
371 self.mk_sp(start, suffix_start),
372 "unterminated character literal",
373 error_code!(E0762),
374 )
375 }
376 (token::Char, Mode::Char, 1, 1) // ' '
377 }
378 rustc_lexer::LiteralKind::Byte { terminated } => {
379 if !terminated {
380 self.sess.span_diagnostic.span_fatal_with_code(
381 self.mk_sp(start + BytePos(1), suffix_start),
382 "unterminated byte constant",
383 error_code!(E0763),
384 )
385 }
386 (token::Byte, Mode::Byte, 2, 1) // b' '
387 }
388 rustc_lexer::LiteralKind::Str { terminated } => {
389 if !terminated {
390 self.sess.span_diagnostic.span_fatal_with_code(
391 self.mk_sp(start, suffix_start),
392 "unterminated double quote string",
393 error_code!(E0765),
394 )
395 }
396 (token::Str, Mode::Str, 1, 1) // " "
397 }
398 rustc_lexer::LiteralKind::ByteStr { terminated } => {
399 if !terminated {
400 self.sess.span_diagnostic.span_fatal_with_code(
401 self.mk_sp(start + BytePos(1), suffix_start),
402 "unterminated double quote byte string",
403 error_code!(E0766),
404 )
405 }
406 (token::ByteStr, Mode::ByteStr, 2, 1) // b" "
407 }
408 rustc_lexer::LiteralKind::RawStr { n_hashes, err } => {
409 self.report_raw_str_error(start, err);
410 let n = u32::from(n_hashes);
411 (token::StrRaw(n_hashes), Mode::RawStr, 2 + n, 1 + n) // r##" "##
412 }
413 rustc_lexer::LiteralKind::RawByteStr { n_hashes, err } => {
414 self.report_raw_str_error(start, err);
415 let n = u32::from(n_hashes);
416 (token::ByteStrRaw(n_hashes), Mode::RawByteStr, 3 + n, 1 + n) // br##" "##
417 }
418 rustc_lexer::LiteralKind::Int { base, empty_int } => {
419 return if empty_int {
420 self.sess
421 .span_diagnostic
422 .struct_span_err_with_code(
423 self.mk_sp(start, suffix_start),
424 "no valid digits found for number",
425 error_code!(E0768),
426 )
427 .emit();
428 (token::Integer, sym::integer(0))
429 } else {
430 self.validate_int_literal(base, start, suffix_start);
431 (token::Integer, self.symbol_from_to(start, suffix_start))
432 };
433 }
434 rustc_lexer::LiteralKind::Float { base, empty_exponent } => {
435 if empty_exponent {
436 self.err_span_(start, self.pos, "expected at least one digit in exponent");
437 }
438
439 match base {
440 Base::Hexadecimal => self.err_span_(
441 start,
442 suffix_start,
443 "hexadecimal float literal is not supported",
444 ),
445 Base::Octal => {
446 self.err_span_(start, suffix_start, "octal float literal is not supported")
447 }
448 Base::Binary => {
449 self.err_span_(start, suffix_start, "binary float literal is not supported")
450 }
451 _ => (),
452 }
453
454 let id = self.symbol_from_to(start, suffix_start);
455 return (token::Float, id);
456 }
457 };
458 let content_start = start + BytePos(prefix_len);
459 let content_end = suffix_start - BytePos(postfix_len);
460 let id = self.symbol_from_to(content_start, content_end);
461 self.validate_literal_escape(mode, content_start, content_end, prefix_len, postfix_len);
462 (lit_kind, id)
463 }
464
465 #[inline]
src_index(&self, pos: BytePos) -> usize466 fn src_index(&self, pos: BytePos) -> usize {
467 (pos - self.start_pos).to_usize()
468 }
469
470 /// Slice of the source text from `start` up to but excluding `self.pos`,
471 /// meaning the slice does not include the character `self.ch`.
str_from(&self, start: BytePos) -> &str472 fn str_from(&self, start: BytePos) -> &str {
473 self.str_from_to(start, self.pos)
474 }
475
476 /// As symbol_from, with an explicit endpoint.
symbol_from_to(&self, start: BytePos, end: BytePos) -> Symbol477 fn symbol_from_to(&self, start: BytePos, end: BytePos) -> Symbol {
478 debug!("taking an ident from {:?} to {:?}", start, end);
479 Symbol::intern(self.str_from_to(start, end))
480 }
481
482 /// Slice of the source text spanning from `start` up to but excluding `end`.
str_from_to(&self, start: BytePos, end: BytePos) -> &str483 fn str_from_to(&self, start: BytePos, end: BytePos) -> &str {
484 &self.src[self.src_index(start)..self.src_index(end)]
485 }
486
report_raw_str_error(&self, start: BytePos, opt_err: Option<RawStrError>)487 fn report_raw_str_error(&self, start: BytePos, opt_err: Option<RawStrError>) {
488 match opt_err {
489 Some(RawStrError::InvalidStarter { bad_char }) => {
490 self.report_non_started_raw_string(start, bad_char)
491 }
492 Some(RawStrError::NoTerminator { expected, found, possible_terminator_offset }) => self
493 .report_unterminated_raw_string(start, expected, possible_terminator_offset, found),
494 Some(RawStrError::TooManyDelimiters { found }) => {
495 self.report_too_many_hashes(start, found)
496 }
497 None => (),
498 }
499 }
500
report_non_started_raw_string(&self, start: BytePos, bad_char: char) -> !501 fn report_non_started_raw_string(&self, start: BytePos, bad_char: char) -> ! {
502 self.struct_fatal_span_char(
503 start,
504 self.pos,
505 "found invalid character; only `#` is allowed in raw string delimitation",
506 bad_char,
507 )
508 .emit();
509 FatalError.raise()
510 }
511
report_unterminated_raw_string( &self, start: BytePos, n_hashes: usize, possible_offset: Option<usize>, found_terminators: usize, ) -> !512 fn report_unterminated_raw_string(
513 &self,
514 start: BytePos,
515 n_hashes: usize,
516 possible_offset: Option<usize>,
517 found_terminators: usize,
518 ) -> ! {
519 let mut err = self.sess.span_diagnostic.struct_span_fatal_with_code(
520 self.mk_sp(start, start),
521 "unterminated raw string",
522 error_code!(E0748),
523 );
524
525 err.span_label(self.mk_sp(start, start), "unterminated raw string");
526
527 if n_hashes > 0 {
528 err.note(&format!(
529 "this raw string should be terminated with `\"{}`",
530 "#".repeat(n_hashes)
531 ));
532 }
533
534 if let Some(possible_offset) = possible_offset {
535 let lo = start + BytePos(possible_offset as u32);
536 let hi = lo + BytePos(found_terminators as u32);
537 let span = self.mk_sp(lo, hi);
538 err.span_suggestion(
539 span,
540 "consider terminating the string here",
541 "#".repeat(n_hashes),
542 Applicability::MaybeIncorrect,
543 );
544 }
545
546 err.emit();
547 FatalError.raise()
548 }
549
550 // RFC 3101 introduced the idea of (reserved) prefixes. As of Rust 2021,
551 // using a (unknown) prefix is an error. In earlier editions, however, they
552 // only result in a (allowed by default) lint, and are treated as regular
553 // identifier tokens.
report_unknown_prefix(&self, start: BytePos)554 fn report_unknown_prefix(&self, start: BytePos) {
555 let prefix_span = self.mk_sp(start, self.pos);
556 let prefix_str = self.str_from_to(start, self.pos);
557 let msg = format!("prefix `{}` is unknown", prefix_str);
558
559 let expn_data = prefix_span.ctxt().outer_expn_data();
560
561 if expn_data.edition >= Edition::Edition2021 {
562 // In Rust 2021, this is a hard error.
563 let mut err = self.sess.span_diagnostic.struct_span_err(prefix_span, &msg);
564 err.span_label(prefix_span, "unknown prefix");
565 if prefix_str == "rb" {
566 err.span_suggestion_verbose(
567 prefix_span,
568 "use `br` for a raw byte string",
569 "br".to_string(),
570 Applicability::MaybeIncorrect,
571 );
572 } else if expn_data.is_root() {
573 err.span_suggestion_verbose(
574 prefix_span.shrink_to_hi(),
575 "consider inserting whitespace here",
576 " ".into(),
577 Applicability::MaybeIncorrect,
578 );
579 }
580 err.note("prefixed identifiers and literals are reserved since Rust 2021");
581 err.emit();
582 } else {
583 // Before Rust 2021, only emit a lint for migration.
584 self.sess.buffer_lint_with_diagnostic(
585 &RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX,
586 prefix_span,
587 ast::CRATE_NODE_ID,
588 &msg,
589 BuiltinLintDiagnostics::ReservedPrefix(prefix_span),
590 );
591 }
592 }
593
594 /// Note: It was decided to not add a test case, because it would be too big.
595 /// <https://github.com/rust-lang/rust/pull/50296#issuecomment-392135180>
report_too_many_hashes(&self, start: BytePos, found: usize) -> !596 fn report_too_many_hashes(&self, start: BytePos, found: usize) -> ! {
597 self.fatal_span_(
598 start,
599 self.pos,
600 &format!(
601 "too many `#` symbols: raw strings may be delimited \
602 by up to 65535 `#` symbols, but found {}",
603 found
604 ),
605 )
606 .raise();
607 }
608
validate_literal_escape( &self, mode: Mode, content_start: BytePos, content_end: BytePos, prefix_len: u32, postfix_len: u32, )609 fn validate_literal_escape(
610 &self,
611 mode: Mode,
612 content_start: BytePos,
613 content_end: BytePos,
614 prefix_len: u32,
615 postfix_len: u32,
616 ) {
617 let lit_content = self.str_from_to(content_start, content_end);
618 unescape::unescape_literal(lit_content, mode, &mut |range, result| {
619 // Here we only check for errors. The actual unescaping is done later.
620 if let Err(err) = result {
621 let span_with_quotes = self
622 .mk_sp(content_start - BytePos(prefix_len), content_end + BytePos(postfix_len));
623 let (start, end) = (range.start as u32, range.end as u32);
624 let lo = content_start + BytePos(start);
625 let hi = lo + BytePos(end - start);
626 let span = self.mk_sp(lo, hi);
627 emit_unescape_error(
628 &self.sess.span_diagnostic,
629 lit_content,
630 span_with_quotes,
631 span,
632 mode,
633 range,
634 err,
635 );
636 }
637 });
638 }
639
validate_int_literal(&self, base: Base, content_start: BytePos, content_end: BytePos)640 fn validate_int_literal(&self, base: Base, content_start: BytePos, content_end: BytePos) {
641 let base = match base {
642 Base::Binary => 2,
643 Base::Octal => 8,
644 _ => return,
645 };
646 let s = self.str_from_to(content_start + BytePos(2), content_end);
647 for (idx, c) in s.char_indices() {
648 let idx = idx as u32;
649 if c != '_' && c.to_digit(base).is_none() {
650 let lo = content_start + BytePos(2 + idx);
651 let hi = content_start + BytePos(2 + idx + c.len_utf8() as u32);
652 self.err_span_(lo, hi, &format!("invalid digit for a base {} literal", base));
653 }
654 }
655 }
656 }
657
nfc_normalize(string: &str) -> Symbol658 pub fn nfc_normalize(string: &str) -> Symbol {
659 use unicode_normalization::{is_nfc_quick, IsNormalized, UnicodeNormalization};
660 match is_nfc_quick(string.chars()) {
661 IsNormalized::Yes => Symbol::intern(string),
662 _ => {
663 let normalized_str: String = string.chars().nfc().collect();
664 Symbol::intern(&normalized_str)
665 }
666 }
667 }
668